The long-term goal of our research program is to develop new treatments for improving behavioral deficits post-stroke based on a formal theory of brain function and organization. However, the immediate goal of this competitive renewal is to identify patterns of disrupted structural and functional connectivity that are related to stroke-induced deficits, determine how they change during recovery, and see if they predict outcome. Computational theories have proposed that lesions to the brain will have different effects based on the underlying network architecture. Mechanisms controlling recovery of function may similarly depend on network architecture. Specific Aim 1 measures changes in whole brain functional connectivity (FC) after damage to peripheral networks that are mainly connected within the network vs. central networks that communicate more broadly with other brain networks. We measure separately the effect of cortical vs. white matter pathway damage. We predict that peripheral lesions (either to gray or white matter) will produce FC changes that are mainly limited to the affected network, while central lesions (either gray or white matter) will produce multi- network FC changes. Peripheral networks are connected to input/output pathways and mediate predominantly sensory or motor functions. Central networks are connected through association pathways and mediate predominantly attention and memory functions. Specific Aim 2 uses a machine learning ridge regression method to test the idea that sensory-motor deficits and their recovery are more dependent on structural damage and anatomical disconnection, whereas cognitive impairments are better captured by multi-network functional connectivity dysfunction. An important clinical goal of the project is to determine whether advanced neuroimaging techniques, such as structural imaging, DTI, and fMRI can provide useful clinical information on an individual patient's recovery and outcome, over and above that provided by measures of acute behavior, demographic variables, and amount of treatment. In Specific Aim 3 we enter neuroimaging data into a classifier to predict outcome in individual patients, as well as to predict the effect of rehabilitation interventions. Finally, in Specific Aim 4 we move from correlation and prediction to mechanism. Our basic hypothesis is that changes in functional connectivity after focal injury reflect alterations in brain dynamics, specifically a decrease in the variability of inter-regional phase differences. This hypothesis is tested empirically by phase measurements, but also computationally, using a biophysically based model that can simulate the FC changes observed in patients. Importantly, the model also estimates the information capacity of the damaged brain, which we will correlate with the patient's empirically measured degree of modularity and profile of behavioral deficits.